The pace of artificial intelligence seems blinding fast as OpenAI announced a significant upgrade to ChatGPT hours ago. It is called GPT-4 and has more features than the previous GPT-3.5. As a result, you can do more with ChatGPT.
For example, it can respond to images. Upload a picture, and it can describe the objects shown. Consequently, you can command the AI bot to create content related to specific parts of a photo. It is so advanced that it can write code for a website from a drawing!
ChatGPT-4 is only available in the ChatGPT Plus program, which costs $20 monthly. Alternatively, you may try the latest large language model on the Bing AI chatbot for free. Let us discuss the features of ChatGPT-4 so that you can see whether it is worth trying.
What are the features of ChatGPT-4?
Some people assume ChatGPT and GPT are the same. However, the former refers to the chatbot that lets them use the GPT large language model.
That means GPT got the upgrade, not ChatGPT. Still, I would sometimes refer to GPT-4 as “ChatGPT-4” to avoid repeating the former excessively throughout this article. Here are the new things GPT-4 can do:
- GPT-4 can see and understand photos.
- GPT-4 has different “personalities.”
- ChatGPT-4 is harder to trick or “jailbreak.”
- ChatGPT-4 has a longer memory.
- GPT-4 works with more languages.
1. GPT-4 can see and understand photos
Perhaps the most impressive change to ChatGPT is that it can accept image queries. For example, let’s say you sent it a photo of a yellow iPhone 14.
Then, you can ask questions about the image, such as “What are its features” or “How much does it cost.” In response, it would list a comprehensive answer to your question.
Moreover, it could act as your translator, quickly translating long foreign texts. You could also take a photo of an entire essay, and it would turn it into text you can copy and paste.
You may also like: Meet DALL-E, The Text-To-Image Generator
GPT-4 can accept, remember, and act on over 20,000 words. That means you can take a photo of a scientific journal or an academic paper and immediately receive a summary.
Note that I only mentioned that as an example. This article does not condone using AI for cheating. Please follow your school’s rules when using this tool.
2. GPT-4 has different “personalities.”
Most people don’t want to receive an entire paragraph when asking questions. Sometimes, you may prefer simple and casual answers to queries.
Nowadays, AI researchers are exploring a concept called “steerability,” the ability of artificial intelligence to change its behavior on demand.
ChatGPT founder OpenAI enabled GPT-4 to take advantage of steerability by letting users decide whether the chatbot would answer in a casual or professional tone.
You may also like: The Top 10 Applications of ChatGPT In Everyday Life
Moreover, you can ask ChatGPT-4 to roleplay. For example, you can ask it to “answer as if you are a person being interviewed for cable news.”
Microsoft invested heavily in OpenAI, so it applied ChatGPT to its Bing search engine weeks ago. Back then, it let people choose conversational styles.
Bing AI also had a secret “celebrity mode” that allowed it to speak like famous people. It turns out it was already using GPT-4 before its official release!
3. ChatGPT-4 is harder to trick or “jailbreak.”
GPT-3.5 had always been able to roleplay, but people exploited that feature to make the AI answer explicit and harmful content. For example, they want the chatbot to say curse words.
ChatGPT’s limitations prevent them from doing so, but they would break those rules to play user-given roles. The most common method is to ask ChatGPT to act as if it was a person who would say expletives.
The previous version would cause the bot to disobey the built-in guidelines to follow a user’s request. In contrast, ChatGPT-4 will politely notify the user that it may not perform the user’s command.
4. ChatGPT-4 has a longer memory
ChatGPT cannot remember all your prompts, so it will forget some of them while you’re using the tool. Consequently, it may not remember something you shared a few lines of text ago.
GPT-3.5 can only remember around 8,000 words, or around four to five book pages. In contrast, GPT-4 can keep track of 50 pages of text or roughly 64,000 words.
You may also like: 10 Ways To Make Money With ChatGPT
Remember what I said about feeding it an entire journal via photo? The old version can’t do that, but the new ChatGPT-4 can. As a result, it can remember parts of a long text, letting you search for specific segments.
5. GPT-4 works with more languages
OpenAI’s headquarters is in San Francisco, California, so it makes sense for its ChatGPT to focus on English. Still, it had limited support for other languages.
For example, the old and new versions can answer you in Tagalog. Nowadays, GPT-4 can provide multiple-choice answers for 26 languages, such as Korean and Ukrainian.
ChatGPT-4 is not foolproof, especially for languages that use different characters instead of the English alphabet. It can still make errors that you must spot and correct yourself.
Conclusion
OpenAI continues to lead AI chatbot development with the release of GPT-4. Note that it is only available for ChatGPT Plus subscribers, and the subscription costs $20 monthly.
Fortunately, OpenAI expanded access to customers outside the United States, so you could start using it for $20 monthly. On the other hand, you can use a free limited version via Bing AI.
Do you want to learn more about the exciting developments in AI technology? Follow the latest trends in artificial intelligence, virtual reality, and more at Inquirer Tech.
Frequently asked questions about ChatGPT-4
Can I use ChatGPT-4 for free?
At the time of writing, GPT-4 is only available to ChatGPT Plus subscribers, and the subscription costs $20 monthly. On the other hand, you could use a free version by accessing the Bing AI chatbot. Note that it offers a limited number of prompts per user, so you may not use it for long periods. You may consider it a “free trial” for GPT-4.
What is the difference between ChatGPT-4 and 3?
The previous version was GPT-3.5, which was unimodal. In other words, it could only accept and process text prompts, such as essays or questions. On the other hand, the latest version, called GPT-4, is multimodal. It can accept images, allowing it to provide descriptions of specific objects within photos.
Is GPT-4 the most powerful AI?
At the time of writing, GPT-4 is one of the most powerful known artificial intelligence systems. It can accept, remember, and act on over 20,000 words. As a result, you could provide photos of scientific journals or academic papers and immediately receive a summary. However, OpenAI may release an even more powerful AI in December called GPT-5.